datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
pretraining_v1-omega_booksAI-CUDA-Engineer-Archive
The AI CUDA Engineer Archive 👷: Agentic CUDA Kernel Discovery, Optimization & Composition
We release The AI CUDA Engineer archive, a dataset consisting of approximately 30,000 CUDA kernels generated by The AI CUDA Engineer. It is released under the CC-By-4.0 license and can be accessed via HuggingFace and interactively visualized here. The dataset is based on the Kernel tasks provided in KernelBench and includes a torch reference implementation, torch, NCU and Clang-tidy… See the full description on the dataset page: https://huggingface.co/datasets/SakanaAI/AI-CUDA-Engineer-Archive.L2DTL;DR of L2D, the world's largest self-driving dataset! Read more about L2D on the official Huggingface blog: LeRobot goes to driving school
90+ TeraBytes of multimodal data (5000+ hours of driving) from 30 cities in Germany
6x surrounding HD cameras and complete vehicle state: Speed/Heading/GPS/IMU
Continuous: Gas/Brake/Steering and discrete actions: Gear/Turn Signals
Environment state: Lane count, Road type (highway|residential), Road surface (asphalt, cobbled, sett), Max speed limit.… See the full description on the dataset page: https://huggingface.co/datasets/yaak-ai/L2D.cad-environments
CAD Environments
CAD Environments is a multimodal dataset of complete, human-performed workflows in desktop CAD software. The current release contains 51 task workflows totaling 99.03 hours, covering eight software groups across mechanical design, architecture, MEP, structural design, and general 3D modeling.
Each workflow preserves the full task context—not just the final model—including the problem statement, reference and input files, a gold output, evaluation rubrics, a… See the full description on the dataset page: https://huggingface.co/datasets/markov-ai/cad-environments.computer-use-large
Computer Use Large
A large-scale dataset of 48,478 screen recording videos (~12,300 hours) of professional software being used, sourced from the internet. All videos have been trimmed to remove non-screen-recording content (intros, outros, talking heads, transitions) and audio has been stripped.
Dataset Summary
Category
Videos
Hours
AutoCAD
10,059
2,149
Blender
11,493
3,624
Excel
8,111
2,002
Photoshop
10,704
2,060
Salesforce
7,807
2,336
VS… See the full description on the dataset page: https://huggingface.co/datasets/markov-ai/computer-use-large.fineweb-edu-fortified
Fineweb-Edu-Fortified
The composition of fineweb-edu-fortified, produced by automatically clustering a 500k row sample in
Airtrain
What is it?
Fineweb-Edu-Fortified is a dataset derived from
Fineweb-Edu by applying exact-match
deduplication across the whole dataset and producing an embedding for each row. The number of times
the text from each row appears is also included as a count column. The embeddings were produced
using TaylorAI/bge-micro
Fineweb and… See the full description on the dataset page: https://huggingface.co/datasets/airtrain-ai/fineweb-edu-fortified.leaderboard-dataset
Arena Leaderboard Dataset
Historical snapshots of the Arena leaderboard.
Usage
from datasets import load_dataset
# Load all historical text style control data
ds = load_dataset("lmarena-ai/leaderboard-dataset", "text_style_control", split="full")
# Load the current text style control leaderboard
ds = load_dataset("lmarena-ai/leaderboard-dataset", "text_style_control", split="latest")
# Filter to overall category
ds =… See the full description on the dataset page: https://huggingface.co/datasets/lmarena-ai/leaderboard-dataset.aime_2025
AIME 2025
This dataset contains 30 problems from the 2025 AIME tests, including:
AIME I: 15 problems
AIME II: 15 problems
airlens-live
AirLens Live Data
Live data layer for AirLens, an open
air-quality monitoring platform. Updated by scheduled GitHub Actions pipelines.
Layout mirrors the former Supabase Storage buckets:
Path
Content
Cadence
aq-data/current-*-grid.json
Global pollutant grids (PM2.5/PM10/O3/NO2/CO)
hourly
aq-data/timeline/
GEFS-Aerosols PM2.5 frames, -24h..+24h, 3h step
every 3h
aq-data/predictions/grid_latest.json
AOD→PM2.5 model predictions (p10-p90 + DQSS)
every 3h… See the full description on the dataset page: https://huggingface.co/datasets/Robeedau/airlens-live.aime_2026
Homepage and repository
Homepage: https://matharena.ai/
Repository: https://github.com/eth-sri/matharena
Dataset Summary
This dataset contains the questions from AIME 2026 used for the MathArena Leaderboard
Data Fields
The dataset contains the following fields:
problem_idx (int64): Problem index within the corresponding MathArena benchmark.
answer (int64): Gold final answer.
problem (string): Problem statement, usually stored as LaTeX source.
Source… See the full description on the dataset page: https://huggingface.co/datasets/MathArena/aime_2026.aime_2025
Homepage and repository
Homepage: https://matharena.ai/
Repository: https://github.com/eth-sri/matharena
Dataset Summary
This dataset contains the questions from AIME 2025 used for the MathArena Leaderboard
Data Fields
The dataset contains the following fields:
problem_idx (int64): Problem index within the corresponding MathArena benchmark.
problem (string): Problem statement, usually stored as LaTeX source.
answer (int64): Gold final answer.
problem_type… See the full description on the dataset page: https://huggingface.co/datasets/MathArena/aime_2025.MOSAIC-Refactoring
Agentic Pull Request Dataset
Dataset Overview
The dataset contains 4,910,698 Pull Requests in total, consisting of 4,392,818 agent-authored PRs from 10 agents and 517,880 human-authored PRs. The agent-authored PRs come from Claude, Codegen, Codex, Copilot, Cosine, Cursor, Devin, Jules, Junie, and OpenHands. A summary of the dataset is presented below.
Cohort
Pull Requests
Merged Pull Requests
Repositories
Sum of Additions
Sum of Deletions
Humans
517880… See the full description on the dataset page: https://huggingface.co/datasets/AISE-TUDelft/MOSAIC-Refactoring.AIME_1983_2024Disclaimer: This is a Benchmark dataset! Do not using in training!
This is the Benchmark of AIME from year 1983~2023, and 2024(part 2).
Original: https://artofproblemsolving.com/wiki/index.php/AIME_Problems_and_Solutions
2024(part 1) can be find at https://huggingface.co/datasets/AI-MO/aimo-validation-aime.
Citation
@misc {di_zhang_2025,
author = { {Di Zhang} },
title = { AIME_1983_2024 (Revision 6283828) },
year = 2025,
url = {… See the full description on the dataset page: https://huggingface.co/datasets/di-zhang-fdu/AIME_1983_2024.arc-whestbench-public-2026
Organized by:
Alignment Research Center (ARC),
AIcrowd
WhestBench 2026: ARC White-Box Estimation Challenge
WhestBench is a benchmark for white-box activation estimation: given the weights of a randomly initialized ReLU multi-layer perceptron (MLP) and a strict floating-point-operation (FLOP) budget, predict the average post-activation value of every neuron when the network is fed standard Gaussian inputs.
This is the WhestBench 2026… See the full description on the dataset page: https://huggingface.co/datasets/aicrowd/arc-whestbench-public-2026.Aiice
Dataset
Aiice benchmark dataset for Arctic sea ice concentration (SIC) forecasting,
based on OSI-SAF satellite products (CC BY 4.0).
Coverage
Period: October 1978 – April 2026
Resolution: 25 km spatial, daily temporal
Grid: 432×432 (Lambert Azimuthal Equal Area, EPSG:6931)
Source products
Product
Source
Period
OSI-450-a
SMMR, SSM/I, SSMIS
1978–2020
OSI-430-a
SSMIS
2021–Jul 2025
OSI-438
AMSR2
Jul 2025–present… See the full description on the dataset page: https://huggingface.co/datasets/ITMO-NSS/Aiice.ai-model-popularity
Datamata AI Model Popularity Index
Weekly popularity of the most-downloaded and trending Hugging Face models: trailing downloads, likes, the model's task and its trending rank. One row per model from the most recent weekly snapshot.
Latest snapshot: 2026-09-20
Models in this release: 50
Updated: weekly
Licence: CC BY 4.0 — free to use and adapt, including commercially, with attribution.
Source & methodology: https://www.datamatastudios.com/datasets
Quickstart… See the full description on the dataset page: https://huggingface.co/datasets/datamatastudios/ai-model-popularity.cad-gen-freecad-bench
Parametric CAD Bench — results dataset
Run-by-run results for Parametric CAD Bench, a benchmark that
measures whether AI agents can author editable FreeCAD models from
natural-language part descriptions. 1000 rows, one per
(agent, model, task_id, trial) over the
gnucleus-ai/cad-bench@v1
task suite. The public leaderboard view of this data lives at
cadbench.ai.
What's in here
data/cad-bench-v1.parquet — the row table. Each row carries the
composite + sub-scores… See the full description on the dataset page: https://huggingface.co/datasets/gnucleus-ai/cad-gen-freecad-bench.nexar_collision_prediction
Nexar Collision Prediction Dataset
This dataset is part of the Nexar Dashcam Crash Prediction Challenge on Kaggle.
Dataset
The Nexar collision prediction dataset comprises videos from Nexar dashcams. Videos have a resolution of 1280x720 at 30 frames per second and typically have about 40 seconds of duration. The dataset contains 1500 videos where half show events where there was a collision or a collision was eminent (positive cases), and the other half shows regular… See the full description on the dataset page: https://huggingface.co/datasets/nexar-ai/nexar_collision_prediction.aimo-validation-amc
Dataset Card for AIMO Validation AMC
All 83 come from AMC12 2022, AMC12 2023, and have been extracted from the AOPS wiki page https://artofproblemsolving.com/wiki/index.php/AMC_12_Problems_and_Solutions
This dataset serves as an internal validation set during our participation in the AIMO progress prize competition. Using data after 2021 is to avoid potential overlap with the MATH training set.
Here are the different columns in the dataset:
problem: the modified problem statement… See the full description on the dataset page: https://huggingface.co/datasets/AI-MO/aimo-validation-amc.UltraVideo
UltraVideo: High-Quality UHD 4K Video Dataset
🤓 Project | 📑 Paper | 🤗 Hugging Face (UltraVideo Dataset)) | 🤗 Hugging Face (UltraVideo-Long Dataset)) | 🤗 Hugging Face (UltraWan-1K/4K Weights)
UltraVideo: High-Quality UHD Video Dataset with Comprehensive Captions
🎋 Click below image to watch the 4K demo video.
🤓 First open-sourced UHD-4K/8K video datasets with comprehensive structured (10 types) captions.🤓 Native 1K/4K videos generation by UltraWan.… See the full description on the dataset page: https://huggingface.co/datasets/APRIL-AIGC/UltraVideo.gaming-500-hours
Gaming Dataset (gaming-1) — 494.7 Hours
Native PC/console gameplay screen-recordings, organized by game. Each workflow
is one play session, trimmed to pure gameplay — login screens, launchers,
desktop, collection-app references, and any watching/streaming are removed.
In-game menus, lobbies, loading, and cutscenes are retained as part of the session.
Workflows: 776
Total gameplay: 494.7 hours
Distinct games: 168
Clip duration (min): median 24.0, p90 90.9, max 457.7
Platforms:… See the full description on the dataset page: https://huggingface.co/datasets/markov-ai/gaming-500-hours.multimodal-ct-radiology-reports
Perle AI Multi-phase CECT and CT with Radiology Reports
Summary
A de-identified CT dataset from Perle AI, paired with the original radiology reports. It supports work on multi-modal medical imaging: phase or pathology classification, report generation from images, and visual question answering.
The release has three configurations:
Config
Modality
Subjects
Pairing
cect_3phase
3-phase contrast-enhanced abdominal CT (DICOM)
5
per-subject text report +… See the full description on the dataset page: https://huggingface.co/datasets/Perle-ai/multimodal-ct-radiology-reports.thai-commoncrawl-index
Thai Common Crawl Index (2019–2026)
An index of every page Common Crawl detected as Thai across 70 monthly crawls, from
January 2019 (CC-MAIN-2019-04) to August 2026 (CC-MAIN-2026-30).
932,874,727 page captures · 450,971,497 unique URLs · 6,997,185 hosts · 6,674,969 domains
Each row records where the page lives inside Common Crawl's WARC archives — file name,
byte offset, and record length — so you can fetch exactly the pages you want with HTTP
range requests, without scanning… See the full description on the dataset page: https://huggingface.co/datasets/wayu-ai/thai-commoncrawl-index.TemplateGSM
TemplateMath: Template-based Data Generation (TDG)
This is the official repository for the paper "Training and Evaluating Language Models with Template-based Data Generation", published at the ICLR 2025 DATA-FM Workshop.
Our work introduces Template-based Data Generation (TDG), a scalable paradigm to address the critical data bottleneck in training LLMs for complex reasoning tasks. We use TDG to create TemplateGSM, a massive dataset designed to unlock the next level of… See the full description on the dataset page: https://huggingface.co/datasets/math-ai/TemplateGSM.aitw-processed-labeled-full
AiTW Processed Full with App Labels
This repository contains a full processed Android in the Wild (AiTW) mirror together with an app-labeled step index, official split assignment by episode_id, major-app statistics, and a ready-to-train Gmail subset.
Why This Exists
AiTW is large and not easy to navigate by app. The original labels contain useful fields such as goal_info, current_activity, and action coordinates, but users often need extra processing before they… See the full description on the dataset page: https://huggingface.co/datasets/KMK040412/aitw-processed-labeled-full.the-heap
The Heap Dataset
We develop The Heap, a new contamination-free multilingual code dataset comprising 57 languages, which facilitates LLM evaluation reproducibility. The reproduction packge can be found here.
Is your code in The Heap?
If you would like to have your data removed from the dataset, follow the instructions on GitHub.
Citation
If you use this dataset as part of your research please cite us:
@INPROCEEDINGS {11052803,
author = { Katzy, Jonathan and… See the full description on the dataset page: https://huggingface.co/datasets/AISE-TUDelft/the-heap.epic-kitchens-100-clips
EPIC-KITCHENS-100 Extracted Clips
About
Dataset of 37455 video clips (24GB) extracted from videos in the EPIC-KITCHENS-100 dataset,
more precisely the extension part not contained in EPIC-KITCHENS-55. For details,
see https://www.lightly.ai/product-updates/epickitchens-100-in-lightlystudio.
The clips folder contains one video for every narration from action annotations stored
in {participant_id}/{narration_id}.mp4. The videos have been downscaled an compressed for easier… See the full description on the dataset page: https://huggingface.co/datasets/lightly-ai/epic-kitchens-100-clips.worldcup2026
⚽ WorldCup Arena
A Leakage-Free Forecasting Benchmark on a Live Tournament
Can a language model forecast a match — when the match had not been played at the moment it was asked?
🌐 Language / 语言 : 中文 ▾
📊 四张表
点开本页顶部的 Data Studio 标签即可浏览，也可以直接按名字加载。
Config
行数
内容
fixtures
104
基准本体 —— 喂给模型的头部信息，以及结算后的 90 分钟赛果，七个盘口全部推导好（outcome_1x2、over_2_5、both_score、odd_total）
dossiers
2,208
简报索引 —— 46 快照 × 48… See the full description on the dataset page: https://huggingface.co/datasets/Social-AI-2026/worldcup2026.AIDev
AIDev: Studying AI Coding Agents on GitHub (The Rise of AI Teammates in Software Engineering 3.0)
Papers:
The Rise of AI Teammates in Software Engineering (SE) 3.0: How Autonomous Coding Agents Are Reshaping Software Engineering
AIDev: Studying AI Coding Agents on GitHub
GitHub: https://github.com/SAILResearch/AI_Teammates_in_SE3
This is AIDev v4 (AIDev-2.7M, cutoff date of Nov 2025). Other versions are available
as git tags and can be loaded with… See the full description on the dataset page: https://huggingface.co/datasets/hao-li/AIDev.hplt2_edu_scores
HPLT2-Edu-scores
Dataset summary
HPLT2-JQL-Education is a model-annotated language subset of HPLT2, spanning 35 languages.
Our model-annotations allow for a filtering that achieves higher-quality training outcomes without excessively aggressive data reduction.
The original FW2 heuristic filtering method serves as our baseline, providing reference points for both the volume of retained tokens and downstream model performance.
For example, in the Spanish language case… See the full description on the dataset page: https://huggingface.co/datasets/JQL-AI/hplt2_edu_scores.